Research map / Label-efficient and robust learning / OOD and anomaly detection
PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation
Tal Reiss, Niv Cohen, Liron Bergman, Yedid Hoshen
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Adapt ImageNet-pretrained features with compactness training and elastic regularization | PANDA on CIFAR10 · CIFAR10 | ROC AUC (%): 92.5 -> 96.2 | Improved |
| Adapt ImageNet-pretrained features with compactness training and elastic regularization | PANDA on CatsVsDogs · CatsVsDogs | ROC AUC (%): 96.0 -> 97.3 | Improved |
| Add outlier exposure using 80M Tiny Images | PANDA on CIFAR10 · CIFAR10 | ROC AUC (%): 96.2 -> 98.9 | Improved |
| Add outlier exposure using 80M Tiny Images | PANDA on Fashion MNIST · Fashion MNIST | ROC AUC (%): 95.6 -> 91.8 | Hurt |
| Use pretrained features to initialize RotNet (ablation) | MHRot on CIFAR10 · CIFAR10 | ROC AUC (%): 90.1 -> 86.6 | Hurt |
| Use PANDA-OE rather than MHRot with outlier exposure | CIFAR10 outlier exposure · CIFAR10 | ROC AUC (%): 95.6 -> 98.9 | Improved |
| Use SPADE pretrained multiscale pixel features and nearest-neighbor scoring | MVTec anomaly segmentation · MVTec | pixel-level ROC AUC (%): 89 -> 96.2 | Improved |
| Use sample-wise rather than constant-duration early stopping (ablation) | PANDA on CatsVsDogs · CatsVsDogs | ROC AUC (%): 91.9 -> 95.7 | Improved |
| Use elastic weight consolidation rather than constant-duration early stopping (ablation) | PANDA on CatsVsDogs · CatsVsDogs | ROC AUC (%): 91.9 -> 97.3 | Improved |
| Use elastic weight consolidation rather than constant-duration early stopping (ablation) | PANDA on CIFAR100 · CIFAR100 | ROC AUC (%): 94.8 -> 94.1 | Hurt |
| Finetune all four ResNet blocks instead of blocks 3 and 4 (ablation) | PANDA-EWC on CIFAR10 · CIFAR10 | ROC AUC (%): 96.2 -> 94.9 | Hurt |
| Use only ResNet layer 0 rather than layers 0, 1, and 2 (ablation) | SPADE on MVTec · MVTec | PRO (%): 97.3 -> 90.6 | Hurt |
Later papers that compared with this method
6 later accepted papers report PANDA as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Learning Optimal Prompt Ensemble for Multi-source Visual Prompt Transfer (AAAI 2026 · better on 12 of 14 shared results)
- Fine-Grained Generalization via Structuralizing Concept and Feature Space into Commonality, Specificity and Confounding (AAAI 2026 · better on 3 of 3 shared results)
- Exploiting Structural Consistency of Chest Anatomy for Unsupervised Anomaly Detection in Radiography Images (TPAMI 2024 · better on 2 of 2 shared results)
- GeneralAD: Anomaly Detection Across Domains by Attending to Distorted Features (ECCV 2024 · better on 0 of 1 shared result)
- Mean-Shifted Contrastive Loss for Anomaly Detection (AAAI 2023 · better on 8 of 8 shared results)
- Adversarial Unsupervised Domain Adaptation With Conditional and Label Shift: Infer, Align and Iterate (ICCV 2021 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.86)
- AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios (CVPR 2025 · similarity 0.86)
- ADPretrain: Advancing Industrial Anomaly Detection via Anomaly Representation Pretraining (NeurIPS 2025 · similarity 0.86)
- RealNet: A Feature Selection Network with Realistic Synthetic Anomaly for Anomaly Detection (CVPR 2024 · similarity 0.85)
- A Comprehensive Augmentation Framework for Anomaly Detection (AAAI 2024 · similarity 0.85)
- We Need to Rethink Benchmarking in Anomaly Detection (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
